calc.rv.score: Calculate a relatedness-weighted score for a given rare...

In KinformR: Relationship-Informed Pedigree and Variant Scoring

Calculate a relatedness-weighted score for a given rare variant.

Description

These scores can be used to compare variants of interest within a family.

Usage

calc.rv.score(
  fam.list,
  affected.weight = 1,
  unaffected.weight = 0.5,
  unaffected.max = 8,
  max.err = 4
)

Arguments

fam.list	A list with the names: 'A.c', 'A.i', 'U.c', 'U.i' respectively containing the affected correct, affected incorrect, unaffected correct and unaffected incorrect. - This can be generated with the function: score.variant.status - where each value in the dictionary is a list containing the reference and the reference's relatives as encoded based on their degree of relatedness to the reference (reference = 0, sibling/parent/offspring = 1, uncle/grandparent = 2, cousin = 3, etc.)
affected.weight	A coefficient to multiply the calculated A.c and A.i relatedness values by.
unaffected.weight	A coefficient to multiply the U.c and U.i relatedness values by.
unaffected.max	This param controls the score given to a first degree unaffected relatives scores decay from this specified maximum by half for each subsequent relationship degree.
max.err	A heuristic cap of the number of incorrect assignments allowed when scoring. When the total number of incorrect (sum of affected and unaffected) is exceeded, the variant's score is set to 0, regardless of the number of points for or against. This simplifies scoring and allows for fast filtering of poor quality variants. Default is 4.

Details

For each encoded relationship, a relationship-informed weight is applied to their sharing or not sharing of a variant. The score for affected status is: (1 / coefficient.of.relatedness) * status.weight For example, an affected cousin (encoded as a 3) would get a score of: (1/0.125) * affected weight 8 * 1 = 8 points in favour of the variant. Whereas for unaffected individuals, scores decay the further a person is in relation to the proband based on the formula: ((unaffected.max2) * coefficient.of.relatedness ) * unaffected.weight For example, with the default unaffected.max of 8. The sister that does not have a variant would get a score of ((82) 0.5) * unaffected.weight (16 * 0.5) * 0.5 = 4 points for the variant. If these were the only two relatives considered we could sum the points and get a score in favour of the variant of 8 + 4 = 12 If there is evidence against a variant, this is factored into the score as: total.score = evidence.for - evidence.against For example, if there were also an affected sibling without the variant we would have the score against of: (1/0.5) * 1 = 2 The final score for the variant would then be for - against = total 12 - 2 = 10 Giving a final score of 10 for the variant. Comparing values across variants can be used to rank them based on pedigree-informed levels of variant sharing across affected and unaffected individuals. Increasing the affected.weight relative to the unaffected.weight will make the scores give more weight to the correct/incorrect status of affected individuals. The default is 2:1 weight for affected relative to unaffected, which accounts for the fact that variants are likely to be incompletely penetrant and we therefore want to be more tolerant of unaffected individuals that have a variant rather than affected individuals that do not.

Input:

Value

A list with three components: score, score.for, score.against.

Examples

relations <- list("A.c" = c(0, 1, 3, 1), "A.i" = c(3), "U.c" = c(1, 2), "U.i" = c(1))
rv.scores <- calc.rv.score(relations)

KinformR documentation built on Feb. 17, 2026, 5:07 p.m.